Author: Uri Blass
Date: 07:13:51 12/15/05
Go up one level in this thread
On December 15, 2005 at 09:34:33, Tord Romstad wrote: >On December 15, 2005 at 06:58:55, Robert Allgeuer wrote: > >>Hi Tord, >>What is ECMGCP? Is it a tactical or a positional test-suite? > >It's purely tactical. It's a set of combinations taken from a book >called "Encyclopaedia of Chess Middlegames", hence the letters "ECM". >Gian-Carlo Pascutto has analysed the positions and removed the most >obvious incorrect solutions, hence the letters "GCP". > >ECMGCP is more or less the standard "quick and dirty" tactical test >suite for computer programs today. It is used in basically the same >way the WAC test suite was used a few years ago. WAC is too easy >on today's hardware; most of the ECMGCP positions are slightly harder. > >>Can you please post the positions. > >Dieter posted the link. > >>PS.: So this also means that new Glaurung has a very high score in ECMGCP as >>well ... > >Previous versions scored well, too. This is a bit strange, and it actually >worries me a bit. Glaurung is usually very good at finding the flashy, >sacrifical tactics which are usually found in books and magazines (the type >you will find in the ECMGCP suite). It is horribly bad at spotting the more >mundane tactics which decide typical computer vs computer games. There >seems to be some sort of serious imbalance in my search, and I suspect that >it hurts the playing strength considerably. I don't understand the reason >for the problem. Glaurung does not use any extensions except the standard >ones (minus the recapture extension, which never worked for me), and only >checks are extended by a full ply. > >Tord I read that Dieter claimed that the ecmgcp test suite contains too many sacrifices and he prefers the arasan test suites so I wonder if you tried to compare results in this test. I wonder if you tried other test suites and I wonder if you tried longer time control. I remember that you had a hash bug in glaurung and it may be interesting to know if fixing that bug has some effect in solving test suites faster and if the effect was bigger at longer time control. Uri
This page took 0 seconds to execute
Last modified: Thu, 15 Apr 21 08:11:13 -0700
Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.